Re: [PATCH] treewide: Flush pcap and log files, if used, before exiting

On Thu, Aug 14, 2025 at 07:12:55AM +0200, Stefano Brivio wrote:

On Thu, 14 Aug 2025 14:10:20 +1000 David Gibson wrote: [snip]

...

Here we're adding a new syscall to work around the problems with _exit(). In which case, why don't we add futex() to the syscall list and go back to exit(3).

Because futex() just came up unexpectedly and Paul and myself had to spend hours figuring that out, and there are good chances we'll get something else like that from glibc in the future.

Yes, but that could happen with literally any libc function, I'm not convinced that exit() is a target that deserves special attention.

On top of that, see CVE-2014-3153 and CVE-2020-14381 about futex().

...

From a quick glance (and intuitively) fsync() is much simpler than that.

Eh. Still not really convinced this isn't more trouble than it's worth, but as you wish.

...

With Laurent working on multi-threading we might well want futexes anyhow.

True, but then I'd still prefer to allow futex() explicitly, rather than re-enabling exit handlers, because that's more predictable.

-- David Gibson (he or they) | I'll have my music baroque, and my code david AT gibson.dropbear.id.au | minimalist, thank you, not the other way | around. http://www.ozlabs.org/~dgibson